Abstract
Using a multidimensional index weighting factors related to income, health, and social mobility—the Index of Deep Disadvantage (IDD)—we rank the well-being of disadvantaged U.S. counties (initial scores below the median IDD) when they were on the cusp of the Great Recession and then again well into the recovery. We compare the characteristics of counties that saw improvements to those that saw declines. We find that a clear majority of counties were stable in relative rank. Counties showing improvement tended to have been worse off prerecession than counties where well-being declined. Improving counties were less likely to be urban, tended to have smaller fractions of the population identifying as Black and larger fractions as white, and had a lower proportion of jobs in manufacturing. Stable counties were, on average, the worst off pre-recession and thus remained the worst off near the end of the recovery. All county groups improved in income and employment through the recovery, but these advances were not consistently associated with gains in other areas such as incidence of low-weight births.
The Great Recession of the late 2000s was the worst and longest-lasting downturn since the Great Depression. It forced many workers into long-term unemployment and others to withdraw from the labor market, and it increased the risk of underemployment for those who remained employed (Grusky, Western, and Wimer 2011; Couch et al. 2018; Kroft et al. 2016). This downturn was followed by what some now call the “long recovery,” which, while protracted, eventually became by some measures the strongest in history. Ultimately, the gains of the recovery even reached less-skilled workers, resulting in earnings growth, declines in poverty and food insecurity, and other positive trends (Ziliak, this volume). The harms of the Great Recession and the benefits of the long recovery did not, however, affect all people and places equally. It remains important to understand the differential consequences of this shock.
In this article, we use a novel index of community-level economic well-being, a time-varying version of the Index of Deep Disadvantage (IDD), to examine the overall trajectories of communities from prior to the Great Recession to well into the subsequent recovery. The IDD is a score, produced using principal component analysis (PCA), that quantifies community conditions using a range of factors, including measures of income, health, and social mobility. It recognizes that communities might experience adverse conditions, or “disadvantage,” across different dimensions, and that examining any one indicator incompletely represents a community’s overall well-being. The IDD allows us to consider whether overall conditions in communities improved or declined during these tumultuous years. We then analyze both the IDD component variables and other measures to identify, first, the geographic, demographic, and economic factors associated with different trajectories; and, second, the nature of change for communities with different trajectories.
We operationalize “community” by using counties as the unit of analysis. County-level data are available for key measures both prior to and following the recession. Community could be defined at still finer units, such as the city/town or the neighborhood, and a wealthy locale and an impoverished locale within the same county might have experienced very different effects from the Great Recession. Data, however, are often limited at these very fine levels during the time period of interest. In much of the United States, counties are an important level of political organization, so our approach makes maximal use of available data at a meaningful level of geography, but with a key limitation. Restricting our sample to counties that were disadvantaged prior to the recession—scoring below the median on our index—we find that a clear majority of counties can be classified as “stable,” moving relatively little in rank over the course of the recession and the recovery. Approximately 17 percent of counties we call “risers” because they moved up the IDD rankings at least one ventile (groups created by rank-ordering the counties by IDD score, then dividing into twenty evenly sized tranches), and approximately 16 percent we call “decliners” because they moved down the rankings at least one ventile. Stable counties that did not change position generally were the most disadvantaged prior to the Great Recession compared to rising and declining counties and, thus, remained the worst off near the end of the recovery.
Although more disadvantaged overall on our initial index, counties that improved from prior to the recession into the recovery were better off than other counties on some of the individual index indicators: they had, on average, lower initial poverty and unemployment rates and higher median income. They also had the largest average proportion of the working-age population (ages 25 to 64) holding a bachelor’s degree or greater, tended to be least reliant on manufacturing as part of their local economies, and were less likely to be urban. Demographically, both declining and rising counties had larger white and smaller Black populations compared to stable counties. Stable counties were also most reliant on manufacturing and tended to be worse off prerecession on virtually all indicators we examine, the sole exception being incidence of low-weight births. Importantly, each county trajectory group saw some benefits from the protracted recovery, with overall declines in unemployment and increases in income. The translation of these gains into other aspects of well-being differs, however.
Background
The Great Recession was unprecedented in the post–Second World War era, with deep unemployment, declines in overall economic activity, losses of income and wealth, and chilling of credit markets, among other consequences. Research finds that not all individuals and households were equally exposed to the hardships of the Great Recession (Bitler and Hoynes 2015; Pfeffer, Danziger, and Schoeni 2013). For instance, the largest proportional wealth losses were concentrated among households of color, and the racial wealth gap not only persisted during the recovery, it grew larger (Weller and Hanks 2018). Addo and Darity (this volume) find that racialized differences in wealth are larger than occupational class differences in wealth, and Black and Latinx households have seen little wealth gains and increased in debt during the recovery.
Other indicators followed similar demographic patterns. Blacks and Hispanics were more likely than whites to experience employment loss during the recession, and for Blacks in particular the probability of reemployment declined (Couch, Fairlie, and Xu 2016). The incomes of Black professional-class workers were particularly affected by the shock of the recession (Biu, Famighetti, and Hamilton, this volume). Using the official poverty measure, the poverty rate among Blacks and Hispanics—already considerably higher than the rate for whites—increased more sharply than it did for whites. By 2010, Blacks had a poverty rate of 23.3 percent and Hispanics 22.4 percent compared to a rate of 9.9 percent for whites (Danziger, Chavez, and Cumberworth 2012).
Outside racial and ethnic disparities in harm from the recession and benefits from the recovery, people with lower levels of education, low incomes, and lesser existing wealth saw larger proportional wealth declines (Pfeffer, Danziger, and Schoeni 2013). Men were hit harder by unemployment than women (Cunningham 2018; Hoynes, Miller, and Schaller 2012) because men are clustered in industries and occupations more likely to be impacted by economic downturns (Hoynes, Miller, and Schaller 2012).
The negative consequences of the recession varied geographically, with some states experiencing surges in unemployment and poverty while others saw modest changes. As an example, in 2006 California’s poverty rate was 12.2 percent and its unemployment rate was 4.9 percent. In 2010, these figures had risen to 16.3 percent and 12.2 percent, respectively. Over that same period, New Hampshire’s poverty rate rose from 5.4 percent to 6.4 percent, and its unemployment rate grew from 3.4 percent to 5.8 percent (University of Kentucky Center for Poverty Research 2020). One correlate of these differences is industrial composition. An analysis of changes in state unemployment rates found that larger shares of gross state product in manufacturing were associated with larger increases in unemployment (Walden 2012).
These state-level patterns were often evident in finer geographic units. Thiede and Monnat (2016) found that counties within some states experienced much larger unemployment impacts than others, with noticeable clustering in parts of the West, the Southeast, and the Midwest. Larger populations of color (percent Hispanic and percent Black), lower prevailing levels of education, and a larger proportion of the local workforce employed in manufacturing and in construction within counties were associated with larger increases in unemployment. In contrast, larger proportions of workers in agriculture, forestry, fishing, hunting, and mining were associated with better outcomes.
Differences in the effects of the recession were found at the city and even the neighborhood level. Examining neighborhoods in Chicago, Williams, Galster, and Verma (2013) find areas of preexisting disadvantage and larger fractions of people of color were more likely to experience declines in local labor and housing markets. Kim and Cubbin (2019) examined Geographic Research on Wellbeing (a survey of postpartum women in California) data and found that previously high-poverty neighborhoods and (paradoxically) majority-white neighborhoods experienced greater economic deterioration. Lerman and Zhang (2012) combined Panel Study of Income Dynamics data with neighborhood-level data on unemployment, poverty, and home prices, finding that high-poverty neighborhood dwellers experienced greater wealth losses and housing challenges than those in low-poverty areas.
Research questions
Studies of community-level recovery from the recession tend to be limited in scope, focusing on particular geographic areas or specific indicators. Existing analyses also end relatively early during the recovery, so it is not possible to infer the complete trajectory of communities from prerecession through the extended but ultimately strong economic rebound. How did community circumstances change, across multiple indicators, from prior to the recession until the end stages of the recovery? How do communities that improved, grew worse off, or remained relatively stable differ?
Methods
This article builds on the “Understanding Communities of Deep Disadvantage” project, an iterative mixed-methods study that seeks to broaden analyses of poverty beyond income-based measures to other dimensions of disadvantage, such as health and economic mobility. The project shifts attention from the individual to the community. It is conducted by an interdisciplinary team of researchers from the University of Michigan and Princeton University (principal investigators H. Luke Shaefer, Kathryn Edin, and Timothy Nelson) with funding from the Robert Wood Johnson Foundation. It uses a combination of big data and systematic, in-depth qualitative interviews and ethnographic observations to better understand communities of deep disadvantage, with the goal of painting a vivid portrait of the lived experiences of individuals and families in the nation’s poorest communities.
Measuring community disadvantage
The first phase of the larger program of study involved the construction of a multidimensional IDD for all counties and the five hundred largest cities in the United States. The IDD draws on census and administrative data to examine vulnerability in three interconnected domains of high salience to Americans: (1) income, using poverty and deep poverty rates that are official metrics of well-being for the nation; (2) health, using life expectancy and low birth weight, both of which are deeply connected to well-being over the life course; and (3) social mobility, using new estimates for counties and cities.
We used PCA to weight these variables. The IDD is the first principal component and accounts for over 60 percent of the variation in the data. It reveals that deep disadvantage across these dimensions is clustered in the United States in the Deep South, the Cotton Belt, Appalachia, the Rio Grande Valley, and across western Native Lands. This article extends the IDD to examine trajectories of communities over time, comparing index ranking as well as its components and other metrics at the cusp of the Great Recession and again well into the recovery.
We begin by constructing two new versions of the IDD, one using only pre-recession data and one using data from as far into the recovery as possible. The original IDD used poverty and deep poverty as measured by the American Community Survey (ACS). However, direct prerecession poverty estimates from the ACS are available only for a subset of U.S. counties (full coverage is available in more recent years), restricting their utility for measurement prior to the downturn. We modify the IDD by using Small Area Income and Poverty Estimates (SAIPE) of county-level poverty rates, which provide coverage of all U.S. counties. The U.S. Census Bureau produces the SAIPE poverty rates using a model estimated from ACS data and predictors including tax filings data and Supplemental Nutrition Assistance Program (SNAP) data (U.S. Census Bureau 2020). Since the SAIPE estimates are model-based predictions, they are subject to uncertainty. We address this limitation by using a three-year average to produce the poverty rates used in the IDD. Prerecession poverty is calculated using the average of the 2004, 2005, and 2006 SAIPE poverty estimates; while postrecession poverty is the average of the 2017, 2018, and 2019 estimates. Deep poverty rates are not available through SAIPE, so we drop this variable from the modified indices.
The postrecession version of our modified index correlates highly (ρ = .89) with the original IDD constructed from ACS five-year poverty estimates and including deep poverty, indicating that our new measure captures much of the variation in the original score despite differences. Note that we use social mobility in calculating our indices measured using Chetty and Hendren’s (2018) estimates of the incomes of individuals whose parents were in the 25th percentile of the national income distribution, but these data are static so not otherwise analyzed.
Sample
We operationalize community using the county as the unit of analysis, balancing small unit size with data availability. Our initial sample includes all U.S. counties (n = 3,141). Our analyses use a sample restricted to counties classified as “disadvantaged,” those below the median value of the IDD prerecession, giving us a subsample of 1,567 U.S. counties.
Analysis
We first divide all counties, including those categorized as advantaged, in our sample into ventiles (twenty evenly sized rank-ordered groups) based on the pre-recession IDD, then repeat the procedure for the postrecession IDD. After restricting to prerecession disadvantaged counties, we categorize counties as “decliners,” “risers,” or “stable,” depending on whether they moved in their ranking of disadvantage relative to other counties. We define a “decliner” as a county that moved beyond the adjacent ventile down the rankings, a “riser” as a county that moved beyond the adjacent ventile up the rankings, and a “stable” county as one that either remained in the same ventile or moved to an adjacent tranche. These procedures evaluate change relative to other counties, not absolute change. A stable county may grow worse on some indicators, for example, but would maintain approximately the same rank if a majority of counties also grew worse on the indicator.
Having assigned counties to “rising,” “declining,” and “stable” categories, we next examine their prerecession characteristics, starting with descriptive statistics on each IDD component. We also consider other factors such as the prerecession unemployment rate and median income, educational attainment as operationalized by the proportion of the working-age population with a bachelor’s degree or greater, racial and ethnic composition, urbanicity, industry mix, and presence of tribal land in the county. We report the F-test from a linear regression, with standard errors clustered by state, to assess whether the differences on these characteristics are statistically significant between groups. After profiling counties in each group, we examine how they changed from pre-recession into the latter stages of the recovery by calculating differences on each component of the IDD and other time-varying indicators. We list variables and their sources in Table 1.
Data and Sources
Results
Initial characteristics of disadvantaged counties
By definition, approximately half of all U.S. counties are categorized as disadvantaged prior to the recession. Approximately 119 million Americans, or 39 percent of the population at that time, resided in disadvantaged counties on the cusp of the downturn. The first column of Table 2 presents descriptive statistics of the prerecession characteristics of counties that we categorize as disadvantaged. The second column presents the correlation with the prerecession IDD score for continuous variables to ascertain the association of the variable with general disadvantage on the cusp of the recession. Note that the IDD is coded such that negative values indicate greater disadvantage and positive values greater advantage. As expected, each of the IDD components is closely related to overall disadvantage, with lower poverty, a lower rate of low-weight births, and a longer life expectancy in more advantaged counties. Advantaged counties had lower pre-recession unemployment rates and higher median income.
Descriptive Statistics of Communities Classified as Disadvantaged prior to the Great Recession and Correlation with Prerecession IDD Score
p < .001.
Demographically, a larger white population was correlated with a greater level of advantage, while a larger Black population was associated with greater disadvantage. Larger Hispanic populations were associated with higher baseline IDD scores, but the correlation is noticeably weaker than for the other two racial-ethnic demographic variables. A larger proportion of the working-age population holding a bachelor’s degree coincides with greater advantage. Industry mix differs across the distribution of the prerecession IDD. Counties more reliant on agriculture experienced greater disadvantage while the pattern was the opposite for construction. Neither manufacturing nor mining was significantly correlated with prerecession disadvantage.
Table 3 presents the relationship between the two categorical variables, whether the county is urban and whether it contains Indigenous tribal land, and the prerecession IDD score. Urban counties were significantly more advantaged on our index measure than were nonurban counties. We did not find a significant difference in disadvantage between counties with and without tribal land, however.
Association of Categorical Variables with Prerecession IDD Score
p < .001.
Change over time
Approximately 16 percent (n = 247) of counties were decliners, dropping more than one ventile in rank from pre- to postrecession. Another 17 percent (n = 263) were risers, increasing in rank more than one ventile. Finally, a clear majority of counties, 67 percent (n = 1,057), were relatively stable in rank, remaining in the same ventile or shifting to an adjacent one. Table 4 presents prerecession descriptive statistics on declining, rising, and stable counties and the F-test from a regression, with standard errors clustered by state, to assess differences between the characteristics of interest among these three groups.
Prerecession Characteristics of Declining, Rising, and Stable Counties
NOTE: F-test is from a linear regression with standard errors clustered by state.
p < .05. **p < .01. ***p < .001.
Decliners tended to be initially more advantaged than risers or stable communities, while communities categorized as stable had the lowest mean IDD score. Thus, the worst-off communities were, on average, the least likely to change ranking, and so remained the worst-off communities near the end of the recovery. We do find differences in the component variables, however. Risers actually had the lowest prerecession poverty rate (15.46; SD = 4.41) and highest life expectancy (76.96 years; SD = 2.05 years) but also the highest rate of low-weight births at 9.77 percent of all births (SD = 1.57 percent). Stable counties, meanwhile, had the highest average poverty rate (19.40; SD = 5.80), were fairly close to risers in incidence of low-weight births (9.57 percent; SD = 1.84 percent), and had the lowest prerecession life expectancy (on average nearly a year lower than decliners and nearly two full years lower than risers at 74.96 years [SD = 1.88 years]).
Similar economic patterns emerge outside the IDD components. Risers had the lowest mean unemployment rate (4.89; SD = 1.57) and the highest median annual income ($46,434, SD = $8,146), while the opposite was true for stable counties, with an average unemployment rate of 6.17 (SD = 1.84) and median income of $43,811 (SD = $7,918). The county trajectory groups differ in other ways. Rising counties tended to have larger average fractions of the working-age population holding a bachelor’s degree (19.13 percent; SD = 7.99 percent) and stable counties the lowest (16.44 percent; SD = 6.83 percent). Demographically, stable counties had, on average, smaller fractions of the population identifying as white (71.81 percent; SD = 22.16 percent) than either decliners (79.33 percent; SD = 19.22 percent) or risers (79.62 percent; SD = 19.67 percent). Stable counties also clearly had the largest mean Black populations (18.06 percent; SD = 19.48 percent), over double that of either decliners (7.49 percent; SD =9.58 percent) or risers (8.52 percent; SD = 13.02 percent). While not statistically significant, stable counties did have a lower fraction of the population identifying as Hispanic (5.89 percent; SD = 13.42 percent) than either of the other two groups (8.62 percent; SD = 17.11 percent for decliners; and 8.78 percent; SD = 15.40 percent for risers). Finally, risers were both less reliant on manufacturing and less likely to be urban than other counties. Slightly more than 9 percent (9.21 percent; SD = 13.43 percent) of all jobs, on average, were in manufacturing in rising counties, compared to 15.81 percent (SD = 13.82 percent) for decliners and 16.79 percent (SD = 13.92 percent) for stable counties.
The map in Figure 1 presents the geographic distribution of rising, declining, and stable counties. The largest group of counties are those identified as advantaged prerecession. While found throughout the country, they are quite clearly predominant in the Northeast, much of the Midwest, and the West. Disadvantaged but stable counties are concentrated in the South, with additional pockets in Montana and the Dakotas, the Southwest, and the upper Midwest (particularly Michigan, but with some in other states). Both rising and declining counties are more sporadically distributed, but some discernable patterns are evident. There is a clear band of rising counties from Texas north into the Dakotas, some in Montana and scattered through other parts of the West and the South. Declining counties are found in the “Rust Belt” region (e.g., Michigan, Illinois, Indiana, Ohio), in some clusters in the South and parts of the Southwest (particularly New Mexico and Nevada), and sporadically elsewhere.
Map of County Trajectories from prior to the Great Recession through the Recovery
Absolute change
Table 5 presents the pre- and postrecession means and standard deviations on our indicators of economic and social well-being for all disadvantaged counties. Poverty was slightly, albeit statistically significantly, higher postrecession, as was the rate of low-weight births. More positively, life expectancy, on average, increased. This finding might be superficially surprising given the widely noted decrease in U.S. life expectancy in the 2010s. Life expectancy in the United States was still growing as of our prerecession time point. It peaked in 2014 and then declined (Woolf and Schoomaker 2019). Thus, our postrecession life expectancy measure, from the 2016 to 2018 average, comes from a period when it was below its peak but still higher than prior to the Great Recession. Outside the IDD components, unemployment declined and median income increased, on average, for our overall sample of disadvantaged counties.
Overall Pre- to Postrecession Change
NOTE: n=1567
p < .01. ***p < .001.
We present trajectories of change for each category of county visually in Figure 2, while Table 6 lists the pre- to postrecession differences on time-varying indicators. The final column of Table 6 presents the F-test from a regression, with state-clustered standard errors, of the prerecession to recovery difference in the characteristic of interest on the indicators for county groups. Differences in change over time were statistically significant on nearly all time-varying factors: the IDD score, poverty, underweight birth rate, life expectancy, and median income. The sole exception to this pattern was unemployment, on which all counties experienced declines of approximately similar magnitude. All saw, on average, more than a 1 percentage point reduction in the unemployment rate from prior to the recession into the late stages of the recovery.
Prerecession to Postrecession Absolute Change by County Trajectory Group
NOTE: n = 1,567. Values are mean and standard deviation of the prerecession to postrecession difference on each variable. F-test is from a linear regression with standard errors clustered by state.
p < .001.
Prerecession and Late Recovery Means of Measures of Community Well-Being by County Trajectory Group
Other than the unemployment rate, the magnitude and, in some cases, direction of change differed across trajectory groups. Declining counties saw an increase in their poverty rate from pre- to postrecession (1.04; SD = 1.83) while risers saw a decrease (–0.66; SD = 1.84). The poverty rate in stable counties actually increased (0.22; SD = 2.00), albeit only slightly relative to the increase that decliners experienced. Among the economic indicators, perhaps the most noticeable shift is the sizable jump in average median income for risers ($5,507; SD = $4,630). Again, though, counties in all trajectory groups experienced average increases in median income. Risers saw a notable rise in life expectancy, an average increase of over a full year (1.64; SD = 2.12). Decliners saw a very slight decrease (–0.06; SD = 1.77) and stable counties a slight increase (0.39; SD = 1.24). The other health indicator, low-weight births, shows stable counties staying approximately the same (0.09; SD = 1.16) while risers saw a noticeable decrease, in this case of slightly less than 2 percentage points (–1.62; SD = 1.45). Decliners saw an increase of more than 1 percentage point (1.35; SD = 1.38).
Discussion
Our clearest finding is that most communities experienced relative stability from the cusp of the Great Recession through the long recovery. The largest group of counties, approximately 67 percent, either remained in the same ventile or moved to one adjacent. Despite the seismic changes of the Great Recession and the boon of the recovery, disadvantaged counties were likely to remain such, while more advantaged counties were similarly likely to remain advantaged. About one-third did noticeably shift relative to other counties, however, with approximately 16 percent declining and 17 percent improving on our IDD composite measure of well-being. Those counties that rose in the rankings tended to have been somewhat worse off prior to the recession on the IDD, while those that remained stable were, on average, the most disadvantaged.
On almost all of our key indicators, stable counties—those that did not change more than the adjacent ventile in the rankings on our IDD composite measure of disadvantage—were on average the worst off prior to the Great Recession. They had lower IDD scores, higher poverty rates, elevated incidence of low birthweight (the rate for rising counties was slightly higher, but neither practically or statistically meaningfully different), higher unemployment rates, lower median income, and lower life expectancy. A possible explanation for the relative stability among many of these counties on the IDD, then, is simply how far removed they were from counties in the central portion of the overall distribution. A county in the ninth ventile is not greatly different from one in the eighth or even the seventh ventile. However, more expansive gaps exist in the lower ventiles, so a more substantial absolute change is required to meaningfully move for the lowest-ranked counties.
Indications of general economic improvement exist across all types of communities by the end of the protracted recovery, including counties that we label as stable and as decliners. Reductions in unemployment and increases in income were the norm across all three trajectory groups. The income gains following the recovery were much larger for the rising counties, but all saw, on average, at least some increase. These improvements were not paralleled by all indicators, however. Decliners grew slightly worse on poverty, for instance, while stable counties were roughly the same pre- and postrecession on that measure. On both median income and life expectancy, the improvements among risers were of noticeably large magnitude compared to changes for other counties. Depending on the measure of well-being used, all geographic areas could appear to have benefited from the long recovery or it could appear that gains were uneven. Still other indicators would show ongoing harm.
The different portrait emerging from various indicators has implications for policy and for research. Monitoring place-based disadvantage through only one lens can be misleading. Based on income and employment alone, all disadvantaged communities eventually benefitted from the long recovery, and, by these metrics, were ultimately better off than they were prior to the Great Recession. Yet examining other indicators, conditions have deteriorated for some counties while markedly improving for others. Changes in other indicators may lag employment and income, but at present some harms of the recession appear to have lingered for a subset of U.S. counties. Place-based interventions to address downturns in the economic cycle should take these differences into account.
In addition to preexisting level of disadvantage, two factors seem to be particularly related to county trajectories from prior to the recession through the recovery: racial and ethnic demographics and industry. Counties with larger Black populations tended to have been among the worst off prior to the recession and moved little in rank over time—essentially, they have benefitted the least from the recovery. In this regard, our community-level findings reflect the unit-level findings of Addo and Darity and of Biu, Famighetti, and Hamilton that appear in this volume. Racial identity remains a defining characteristic of disadvantage in the United States, whether examined at the individual or the community level. It is worth noting again that all trajectory groups, on average, experienced gains in employment and income, so benefits from the much-improved economy of the late recovery did eventually reach these communities even if they were not sufficient to reverse preexisting disparities.
Both declining and stable counties were more reliant on manufacturing as a component of their local economies than rising counties. Manufacturing in the United States generally presented some unusual trends during the recovery, such as a combination of modestly increasing manufacturing employment but declining productivity (Schmalensee 2018). Our findings indicate that, at the community level, the rebound of manufacturing employment did not lead to substantial gains otherwise. The pattern warrants further investigation. Manufacturers may have relocated during this period, for instance, leaving some communities without their economic base. Alternatively, the structure of returns from manufacturing for the communities in which firms are located may have changed, whether through economic or policy shifts.
Implications for the working class
As evidenced by other articles in this volume, no single accepted definition of “the working class” exists, with identifying characteristics based on factors such as level of education, income, form of employment, and combinations thereof. Consequently, no standard indicator of a “working-class county” or any other geographic unit exists. The issue of defining a working-class county is further complicated by within-county heterogeneity. Wayne County, Michigan, which includes industrial cities such as Detroit and Dearborn, for instance, might be considered the archetype of a working-class area. Yet the county also includes affluent and highly educated municipalities such as Grosse Pointe. If we classify a working-class county based on criteria such as a county below the median fraction of the working-age population with a bachelor’s degree or below the U.S. median county income, Wayne County would not be considered working class.
We do find overlaps between notions of the working class and characteristics of our disadvantaged counties. If we use the educational criteria that we described, less than the median proportion of working-age adults holding a bachelor’s degree, to define a working-class county, then about two-thirds of working-class counties would also be considered disadvantaged counties. Conversely, though, to continue the previous example, Wayne County fits our definition of a disadvantaged county but would not meet the criteria for a working-class county.
Setting aside definitional issues, we can extend some of our findings to consider implications for the well-being of members of the working class since the onset of the Great Recession. That is, our disadvantaged counties are the home and work environment for many people who might reasonably be considered “working class,” and changes in community conditions are therefore changes in that environment. Most importantly, since all counties gained, on average, in employment and income, we would expect the economic prospects of many members of the working class to have improved as a result of the recovery. That only a subset of counties appeared to benefit in other ways over this time period, however, is concerning for the working class.
Two predictors of movement up the IDD rankings are related to general notions of working-class status. Aggregate educational attainment was positively associated with rising through the recovery, while greater economic reliance on manufacturing—a traditionally working-class industry—was associated with either stability or decline through this period. The most stereotypical working-class counties in our data gained the least from the protracted recovery, though we again emphasize that defining a county as working class is a challenging task. On the whole, then, some aspects, such as overall increases in income and decreases in unemployment, appear to have improved the contextual conditions for the working class since the Great Recession. Yet the very factors that are typically used to describe the working class, such as education and industry, overlap with the aggregate-level characteristics of counties that experienced only modest gains or even experienced declines on other indicators.
Limitations
We do not analyze the causes of change within communities, which could come from either change in the experiences of people in communities or changing composition due to inflows or outflows. Using county as the unit of analysis could mask important within-county variation, such as a distressed city in an affluent area or a disadvantaged neighborhood within a prosperous municipality. Finally, our approach is purely descriptive and correlational. Factors such as racial and ethnic demographics, local industry mix, and indicators of community health are intertwined in many ways, but we do not investigate these interrelationships here.
Communities in the next crisis?
As of the writing of this article, the United States is in the midst of a new seismic economic upheaval because of the coronavirus pandemic. We cannot say with certainty whether the longer-term trajectories of disadvantage will parallel those of the Great Recession, but our analysis offers some insights into possible consequences of this new shock. Perhaps most importantly, the most vulnerable communities prior to the Great Recession were still among the worst off following the recovery. We have little reason to believe a different outcome is likely for these communities in the latest recession. Analysts have described the initial phases of the economic recovery from the pandemic-related recession as “k-shaped,” with the circumstances of already-advantaged households rebounding quickly and those of disadvantaged households continuing to deteriorate. If this pattern holds and it aggregates to the community level, we might expect to see increasing stratification in indicators of disadvantage.
Service sector jobs and retail jobs have been particularly hard hit due to both policy and behavioral changes in response to the pandemic. Communities reliant on these industries could experience steeper declines in overall well-being, a contrast to the Great Recession when job losses in manufacturing and construction were acute. Noticeably different from the Great Recession is the role of public health measures in response to the pandemic. Some states, for instance, more quickly and extensively encouraged quarantine and curtailed nonessential economic activities. While a short-term shock occurred to communities in states with more immediate and extensive shutdowns, the long-term economic implications and their relationship to successful management of the pandemic are as of yet unclear. Further, relevant geographic differences in public health infrastructure could exist. If wealthy communities, for example, are better able to manage the pandemic, it could influence their trajectory during the economic recovery. On the other hand, the federal government in the CARES Act took a more inclusive approach to income support than in previous recessions, and considerable research suggests that these actions at least in the short-term may have greatly mitigated economic hardship. Understanding how the CARES Act, including a broadly available economic impact payment and greatly expanded unemployment insurance, affected communities differentially will be important.
Conclusion
In the United States, the damage of the Great Recession and the gains of the recovery were unequally distributed. In this article, we used the multifactor IDD to identify disadvantaged counties prior to the recession, defined as scoring below the median on the IDD measure. We then constructed a postrecession IDD. After assigning all U.S. counties to ventiles in the two time periods, we then identified those counties that moved more than the adjacent tranche, terming counties that grew worse off decliners, those that improved as risers, and those that stayed within one ventile of their original position as stable, then compared their characteristics. Counties that were worse off prior to the Great Recession also tended to be ranked quite low following the recovery, and relative stability was the norm despite the magnitude of the crisis and, eventually, the strength of the recovery.
While all counties experienced, on average, improvements on factors such as income and employment, trajectories were mixed on other measures such as poverty and the incidence of low-weight births. Counties with a proportionally larger Black population were both already among the lowest-scoring on the composite index of disadvantage and among those least likely to experience substantial gains. With respect to the working class, two predictors of county trajectories bring pause—counties with higher educational attainment were more likely to rise in the rankings, and counties more reliant on manufacturing were likely to either remain stably ranked or to decline. Two traditional markers of working-class status were associated with counties least likely to benefit from the recovery. Overall, then, our findings give reason for both optimism and concern, particularly as the nation grapples with the economic fallout of the COVID-19 pandemic.
Footnotes
Vincent A. Fusaro is an assistant professor in the Boston College School of Social Work. His research examines the U.S. welfare state for families with a particular focus on state-level policy, including the relationship between state policy differences and material hardships and the influences on state policy design.
H. Luke Shaefer is Hermann and Amalie Kohn Professor of Social Policy and associate dean for research and policy engagement at the Ford School of Public Policy at the University of Michigan. He is also professor of social work and director of Poverty Solutions, an interdisciplinary, presidential initiative.
Jasmine Simington is a PhD candidate in public policy and sociology at the University of Michigan’s Ford School of Public Policy.